AI Use Disclosure

1. Why this document exists

Buyers are right to ask whether an intelligence product is a language model producing plausible text. For MTR the answer is specific and checkable, so it is written down here rather than asserted in a sales conversation.

This disclosure covers the MTR website, delivered MTR Latitude Studies, MTR SIGMA, and the MTR API. For licensed clients it supplements, and does not replace, the AI terms in your agreement.

2. The intelligence itself is not generated

MTR intelligence is produced by a deterministic simulation engine using computational game theory over public data. It is computed, not written. The same inputs produce the same outputs, and every figure is recoverable from the run, model version, and dataset version that produced it.

No language model participates in producing a scored world, a recommended action, a durability score, a support tier, or any other quantitative output. A generative model cannot change a number in a study, because it never touches one.

3. What MTR SIGMA is

MTR SIGMA is a grounded orchestration layer over computed output. It plans what to retrieve, executes retrieval and math against delivered studies, and reports the result with its provenance. Language models are used for understanding the question and composing the reply.

SIGMA is constrained by a grounding contract. Every claim it makes must trace to a specific computed record or to a traversal of the relationship structure connecting those records. A claim that cannot be traced is not made.

4. What MTR SIGMA must refuse

The grounding contract requires refusal rather than approximation. SIGMA is built to say that it cannot answer, in each of the following cases:

• a question about a market, carrier, or year for which nothing has been computed, where it will say so and identify what would need to run;
• a request for an exit or withdrawal probability, where it will return the watchlist framing and state the limitation instead;
• a request to confirm an external claim that no computed evidence supports;
• a request for a financial projection, market share estimate, or other quantity the engine does not produce;
• a request for regulatory, legal, medical, or investment advice.

5. Human judgment

MTR products are built to inform decisions made by people with subject matter expertise. They augment that judgment and do not replace it. Outputs should be reviewed by someone competent to evaluate them before a material decision is taken, which is why the audit chain exists and why weak structural support is reported as weak rather than smoothed.

6. Your inputs, and what happens to them

Different MTR surfaces handle your input differently, and each says which it is. Questions you ask MTR SIGMA are processed on MTR systems: the question is what the agent plans retrieval from, so it necessarily travels to us, and it is logged as described below. The Latitude Simulator is the opposite by design: plan scoring runs in your browser against fields computed in advance, and the plan you mark does not travel to MTR. A statement that nothing reaches MTR is a statement about the Simulator, not about SIGMA.

MTR does not use customer queries or customer-supplied content to train models, and does not permit third-party model providers to train on them.

SIGMA interactions may be logged to operate the service, diagnose faults, and measure quality. Logs are scoped to that purpose, retained no longer than they serve it, and access to them is restricted; the Privacy Policy's client portal section describes the same records.

Because MTR products require no protected health information and no member-level data, you should not submit PHI, personally identifiable information about individuals, or confidential third-party material into a free-text field. If your organization needs to work with such material, it must be authorized in writing first, and it is not required for any product described on this site.

7. Ownership of outputs

Content produced through SIGMA or the API remains subject to the licence and use restrictions applying to the computed intelligence it draws on. No right in the underlying models, methodology, or training material is granted.

MTR may use de-identified, aggregated usage patterns to improve system performance and reliability.

8. Where AI assists in our own writing

MTR uses AI tools internally in the ordinary way a modern company does, including in drafting narrative material and in software development. Where that happens, every quantitative claim in the resulting material still originates in computed output and carries its provenance, and the claim discipline stated in our Terms of Use applies without exception.

9. Limits and liability

No system is free from error. MTR makes no warranty that any AI-assisted interface will be uninterrupted or error-free, and disclaims liability for decisions made in reliance on any output, on the terms set out in our Terms of Use.

MTR may suspend access to an AI interface for a user or organization where usage presents a legal, security, or regulatory risk.

10. Changes and contact

This disclosure will be updated as capabilities, regulation, and controls change, and the revision date above will reflect it. Questions can be sent to legal@mtrgp.com.